Overview

The striking returns from a shorter AAPL drawdown study weaken when the window expands to twenty years. A fixed profit exit can miss long advances without avoiding the deepest decline.

A longer window needs the same rules on both sides

This article was prepared with AI assistance. Every reported backtest figure was checked against frozen data and reproducible scripts. The study covers 5,029 sessions from September 25, 2006, through September 22, 2026. The first ten years select parameters; the next ten keep them fixed and carry positions and pending orders across the boundary.

While in cash, the rule tracks a high in adjusted closing prices. A close below the drawdown threshold triggers a purchase at the next adjusted open. An exit threshold relative to the actual entry price likewise triggers a sale at the next open. Idle cash earns zero. The main results exclude taxes, spreads, and market impact; a separate stress test charges ten basis points per side.

The missing price history was bridged before measuring returns

The AkShare/Sina AAPL series omits all of 2005 and 2006 and the start of 2007. Of the 5,029 sessions in the twenty-year file, 4,908 use AkShare raw prices and a reconstructed multiplicative corporate-action adjustment. A public Yahoo-derived snapshot supplies 119 early gap sessions and two isolated missing sessions; one duplicated US market holiday was removed. The result is an audited mixed-source series, not twenty years of AkShare-only prices.

Across 200 overlapping sessions in 2007, the median relative difference in closing prices was about 0.005%, with a maximum near 0.015%. One daily high on December 14, 2007, differs materially between sources, so high-based trailing-stop comparisons deserve extra caution. The core rule uses close signals and next-open fills.

The training choice did not beat holding in the next decade

Searching integer-point buy drawdowns from 10% to 30% and take-profit levels from 5% to 30% produced a training winner of a 10% drawdown entry and 24% profit exit. Its first-decade CAGR was 25.33%, against 28.09% for buy and hold. If buy and hold is itself an eligible choice, the training period already favors it.

Keeping the 10%/24% rule unchanged for the second decade produced 28.55% CAGR, against 29.24% for buy and hold. Across the full twenty years the figures were 26.91% and 28.65%, respectively. The chart compares the two policies on the same adjusted-price basis.

Annualized AAPL returns for buy and hold versus the 10 percent drawdown and 24 percent exit rule selected in training
The training-selected rule trails buy and hold in the full twenty years and the second decade.

The apparent parameter plateau becomes much narrower

In the earlier five-year holdout, all 30 combinations of 8%-12% entry drawdowns and 20%-25% take-profit levels beat buy and hold. Across the longer 2016-2026 validation decade, only six do. Their median CAGR is 27.87%, below buy and hold at 29.24%. The outlined cell is the 10%/24% training selection.

Heatmap of second-decade CAGR for thirty AAPL drawdown and profit-taking combinations, most below buy and hold
Green cells beat buy and hold; only six of thirty do so.

Across the broader 676-combination grid spanning 5%-30% on both axes, none beats buy and hold over the full twenty years. This is a description of the observed sample, not a probability estimate for future returns.

Costs matter less than the change across years

Charging ten basis points on each purchase and sale, with comparable terminal liquidation, reduces the selected rule's second-decade CAGR to 28.28%, against 29.22% for buy and hold. Costs widen its deficit from 0.69 to about 0.94 percentage points; they do not explain most of the instability.

A separate rolling-year test also varies by period. Three years of training followed by one year of validation yields four wins, eight losses, and one tie in thirteen test years, with median annual excess return of -10.44 percentage points. Five years of training followed by one year produces four wins, four losses, and three ties in eleven tests. These overlapping historical windows are not independent forecast experiments.

Bars showing annual AAPL excess returns after three- and five-year rolling training windows
Neither rolling design delivers consistently positive annual excess returns.

The exit rule did not avoid the hardest drawdown

The simple recovery rule bought in November 2007 and did not exit until October 2009, remaining exposed through the 2008 collapse. The full-period maximum drawdown of the 10%/24% rule and buy and hold was close to -61%. Staged buying and trailing exits changed exposure and risk, but none achieved a higher full-period CAGR in this study.

The earlier reference rule of 10%/23% earned 30.35% in the second decade, slightly above buy and hold. But that setting was adopted after the same decade had been examined; it is not an independently selected first-decade winner. Across the twenty-year split and rolling windows, the evidence is insufficient to treat fixed take-profit trading as a replacement for long-term holding.

This historical simulation is not personalized investment advice. The assumed trading costs are scenarios, while cash yield, taxes, and actual execution impact are not fully included. New data could change the judgment.

Sources

AKShare US stock-history interfaces and adjustment formula: https://akshare.akfamily.xyz/data/stock/stock.html

Apple dividend and stock-split history: https://investor.apple.com/dividend-history/

Public AAPL daily snapshot used for the missing sessions: https://pubmarks.github.io/datasets/stocks/AAPL/ohlcv.csv

Snapshot-generation code and yfinance settings: https://github.com/Pubmarks/datasets/blob/main/scripts/python/yf/lib/ohlcv_fetch.py

Backtest scripts, file hashes, trade logs, and source-disagreement checks are retained in the research workspace. Figures here use the frozen files dated September 24, 2026.